Water resource allocation method and system based on process-by-process multi-model optimization fusion

Abstract

The invention relates to a process-by-process multi-model optimization fusion water resource allocation method and system. The method comprises the steps of obtaining historical water demand and historical water saving potential of a target area based on historical water resource monitoring data; inputting the historical water demand and the historical water-saving potential into a pre-constructed prediction model to obtain a comprehensive water demand and a comprehensive water-saving potential; constructing a three-dimensional joint distribution model based on the historical water resource monitoring data and the water resource management information data, and generating a plurality of water supply scenes based on the three-dimensional joint distribution model; and inputting the water supply scene, the comprehensive water demand and the comprehensive water-saving potential into a pre-constructed optimal configuration model to obtain a water resource configuration optimization scheme. By integrating water regimen, work regimen and other multi-element elements and coupling feedback of a water-saving regulation and control mechanism, a water engineering cluster multi-element joint real-time feedback type scheduling scheme is formulated, and quantitative assessment of water resource utilization and functional perfection of a model in the aspects of hydrological simulation, scheduling, configuration, management and the like are achieved.

Description

Technical Field

[0001] The present application relates to the technical field of water resource allocation, and in particular to a water resource allocation method and system based on process-by-process multi-model optimization fusion. Background Art

[0003] Water resource scheduling model is an important means to achieve efficient management of water conservancy projects. It was originally proposed mainly to address the water competition problem in water-scarce areas. With the deepening of the concept of sustainable development, this issue should be considered not only for water-scarce areas, but also for areas with relatively abundant water resources. From the initial water allocation to the current coordinated consideration of basin and regional economic, environmental and ecological needs for effective water regulation, water resource scheduling research has received increasing attention.

[0004] However, the development and utilization of water resources are increasingly closely integrated with human activities, and the influencing factors are gradually increasing, making its structure more complex. Current research over-generalizes the water sources, users and other factors involved in the water resources system, resulting in calculation results that do not match the actual situation under certain conditions. Therefore, it is necessary to construct a cross-basin water transfer model that can more realistically reflect the actual scheduling requirements, taking into account both simulation and optimization technologies. Summary of the invention

[0005] Based on this, it is necessary to provide a water resources allocation method and system with process-by-process multi-model optimization fusion to address the above-mentioned technical problems and realize the effective mobilization of water resources.

[0006] In a first aspect, the present application provides a water resources allocation method based on process-by-process multi-model optimization fusion.

[0007] The method includes:

[0008] Obtain historical water resources monitoring data and water resources management information data for the target area;

[0009] Based on historical water resources monitoring data, the historical water demand and historical water-saving potential of the target area are obtained; the historical water demand and historical water-saving potential are input into the pre-built prediction model to obtain the comprehensive water demand and comprehensive water-saving potential;

[0010] Based on historical water resources monitoring data and water resources management information data, a three-dimensional joint distribution model is constructed to fit the local water production distribution, water source water regime distribution and upstream water distribution in the target area, and several water supply scenarios are generated based on the three-dimensional joint distribution model;

[0011] The water supply scenario, comprehensive water demand and comprehensive water-saving potential are input into the pre-built optimization configuration model to obtain the optimal water resource allocation plan.

[0012] In one embodiment, the prediction model includes a sub-item trend extrapolation prediction model, a system dynamics model, and a first machine learning model;

[0013] Inputting the historical water demand and historical water-saving potential into a pre-constructed prediction model to obtain the comprehensive water demand and comprehensive water-saving potential includes:

[0014] Inputting the historical water demand and historical water-saving potential into the prediction model to respectively obtain the predicted water demand and predicted water-saving potential corresponding to the sub-item trend extrapolation prediction model, the system dynamics model, and the first machine learning model;

[0015] Based on the historical water demand and historical water-saving potential, using the first optimization algorithm to obtain the first weight values corresponding to the sub-item trend extrapolation prediction model, the system dynamics model, and the first machine learning model;

[0016] Performing weighted fusion on the predicted water demand and predicted water-saving potential based on the first weight values to obtain the comprehensive water demand and comprehensive water-saving potential.

[0017] In one embodiment, a three-dimensional joint distribution model for fitting the local water production distribution, source water condition distribution, and upstream inflow distribution of the target area is constructed based on historical water resource monitoring data and water resource management information data, and several water supply scenarios are generated based on the three-dimensional joint distribution model, including:

[0018] Constructing several water production models for simulating the water production situation of the target area according to the historical water resource monitoring data and water resource management information data, and fusing the water production models to fit the local water production distribution;

[0019] Extracting the source water condition data and upstream inflow data from the historical water resource monitoring data, and fitting the source water condition distribution and upstream inflow distribution according to the source water condition data and upstream inflow data;

[0020] Integrating the local water production distribution, source water condition distribution, and upstream inflow distribution to construct a three-dimensional joint distribution model;

[0021] Performing chaotic sampling based on the three-dimensional joint distribution model to obtain several distribution scenarios;

[0022] Extracting the downstream water demand from the comprehensive water demand, and subtracting the distribution scenarios from the downstream water demand respectively to obtain the corresponding water supply scenarios.

[0023] In one embodiment, the water production model includes a physically based distributed water production model, a conceptual lumped water production model, a statistical model, a rain-flood classification model, and a second machine learning model;

[0024] Fusing the water production models to fit the local water production distribution includes:

[0025] Input the rainfall data of the target area into each water production model to obtain the corresponding local water production data respectively;

[0026] Use the bagging method to average the local water production data corresponding to each water production model to obtain a comprehensive model;

[0027] Input the rainfall data into the comprehensive model, and based on parameter uncertainty, obtain multiple runoff scenarios and fit the distribution of local water production.

[0028] In one embodiment, integrating the local water production distribution, the source water situation distribution, and the upstream inflow distribution, constructing a three-dimensional joint distribution model includes:

[0029] Determine the Markov potential energy functions corresponding to the local water production distribution, the source water situation distribution, and the upstream inflow distribution;

[0030] Convert the Markov potential energy functions into the form of a multi-dimensional joint probability distribution function in the form of a probability distribution;

[0031] Use maximum likelihood estimation to obtain several parameter groups of the Markov potential energy functions; each parameter group estimation includes local potential energy function parameters, pairwise potential energy function parameters, high-order potential energy function parameters, structure parameters, regularization parameters, and constraint parameters;

[0032] Based on each parameter group, construct the corresponding multi-dimensional joint probability distribution function and calculate the information criterion value corresponding to the multi-dimensional joint probability distribution function, compare the information criteria, and determine the three-dimensional joint distribution model from the multi-dimensional joint probability distribution functions according to the comparison results.

[0033] In one embodiment, input the water supply scenario, the comprehensive water demand, and the comprehensive water conservation potential into a pre-constructed optimal allocation model to obtain a water resource allocation optimization plan, including:

[0034] Input the water supply scenario, the comprehensive water demand, and the comprehensive water conservation potential into the pre-constructed optimal allocation model, and use several second optimization algorithms to solve the optimal allocation model to obtain the allocation plans corresponding to each second optimization algorithm; among them, the optimal allocation model takes the minimum total water shortage depth during the water resource scheduling period as the objective function;

[0035] Obtain the historical water supply scenario based on the historical water resource monitoring data, divide the historical water supply scenario, the historical water demand and the historical water conservation potential extracted from the historical water supply scenario into a test data set and a validation data set, input the test data set into the optimal allocation model, use several second optimization algorithms to solve the optimal allocation model to obtain the historical allocation plans corresponding to each second optimization algorithm, and use the validation data set to perform validation calculations on each historical allocation plan to obtain the second weight values corresponding to each optimization algorithm;

[0036] Based on the second weight value, perform weighted fusion on the configuration plan to obtain an optimized water resource allocation plan.

[0037] In one embodiment, the method further includes:

[0038] Construct a generalized river network map of the target area based on the water resource management information data.

[0039] In a second aspect, the present application also provides a water resource allocation system with step-by-step multi-model optimization and fusion. The system includes:

[0040] An acquisition module for obtaining historical water resource monitoring data and water resource management information data of the target area;

[0041] A prediction module for obtaining the historical water demand and historical water-saving potential of the target area based on the historical water resource monitoring data; inputting the historical water demand and historical water-saving potential into a pre-constructed prediction model to obtain the comprehensive water demand and comprehensive water-saving potential;

[0042] A fitting module for constructing a three-dimensional joint distribution model for fitting the local water production distribution, source water condition distribution, and upstream incoming water distribution of the target area based on the historical water resource monitoring data and water resource management information data, and generating a number of water supply scenarios based on the three-dimensional joint distribution model;

[0043] An optimization module for inputting the water supply scenarios, comprehensive water demand, and comprehensive water-saving potential into a pre-constructed optimal allocation model to obtain an optimized water resource allocation plan.

[0044] In a third aspect, the present application also provides a computer device. The computer device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the steps in the above-mentioned water resource allocation method with step-by-step multi-model optimization and fusion.

[0045] In a fourth aspect, the present application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the steps in the above-mentioned water resource allocation method with step-by-step multi-model optimization and fusion.

[0046] In a fifth aspect, the present application also provides a computer program product. The computer program product includes a computer program, and when the computer program is executed by a processor, it implements the steps in the above-mentioned water resource allocation method with step-by-step multi-model optimization and fusion.

[0047] The above-mentioned method and system for optimizing and integrating water resources allocation by process use multi-models to obtain historical water resources monitoring data and water resources management information data of the target area; based on the historical water resources monitoring data, obtain the historical water demand and historical water-saving potential of the target area; input the historical water demand and historical water-saving potential into a pre-constructed prediction model to obtain the comprehensive water demand and comprehensive water-saving potential; construct a three-dimensional joint distribution model for fitting the local water production distribution, source water condition distribution and upstream inflow distribution of the target area based on the historical water resources monitoring data and water resources management information data, and generate several water supply scenarios based on the three-dimensional joint distribution model; input the water supply scenarios, comprehensive water demand and comprehensive water-saving potential into a pre-constructed optimal allocation model to obtain an optimized water resources allocation plan. By integrating multiple elements such as water conditions and engineering conditions, coupling the feedback of the water-saving regulation mechanism, a multi-element joint real-time feedback scheduling plan for water engineering clusters is formulated, the utilization of water resources is quantitatively evaluated, and the dependence on the generalization results of elements such as water sources and users is reduced, so that the water resources allocation plan can better meet the actual scheduling requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 FIG. is a schematic flow chart of a method for optimizing and integrating water resources allocation by process use multi-models in an embodiment;

[0049] Figure 2 FIG. is a schematic flow chart of step S2 in an embodiment;

[0050] Figure 3 FIG. is a schematic flow chart of step S3 in an embodiment;

[0051] Figure 4 FIG. is a schematic flow chart of step S4 in an embodiment;

[0052] Figure 5 FIG. is a structural block diagram of a system for optimizing and integrating water resources allocation by process use multi-models in an embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0053] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0054] An embodiment of the present application provides a method for optimizing and integrating water resources allocation by process use multi-models, as Figure 1 shown, including the following steps:

[0055] S1. Obtain historical water resources monitoring data and water resources management information data of the target area.

[0056] In this embodiment, the target area refers to the area that includes water resource transfer out and transfer in, usually the areas with prominent supply-demand contradictions or uneven water resource distribution. Water resource transfer out areas such as water-rich areas, lake areas, large reservoirs, seasonal surplus areas, etc., and water resource transfer in areas such as arid and semi-arid areas, urban and industrial centers, agricultural areas, seasonal water shortage areas, etc.

[0057] Historical water resource monitoring data includes data such as historical local water production, water source conditions, and upstream inflow, which are collected through water resource monitoring technologies such as hydrological remote sensing technology, flow monitoring technology, water level monitoring technology, etc. Among them, local water production data refers to the quantity and distribution of local natural precipitation, including rainfall, runoff, etc. Water source condition data refers to data such as the water level, flow, and water quality of specific water sources (such as rivers, lakes, reservoirs, etc.). Upstream inflow data refers to the water quantity and water condition information flowing into the target area from the upstream area. Historical water resource monitoring data are all obtained through actual monitoring.

[0058] Water resource management information data includes data reflecting the hydraulic characteristics of river networks such as rivers, lakes, water users, water intake points, cross-sections, nodes, etc., which are obtained through methods such as remote sensing technology, GIS (Geographic Information System), and on-site survey.

[0059] S2. Based on the historical water resource monitoring data, obtain the historical water demand and historical water-saving potential of the target area; input the historical water demand and historical water-saving potential into a pre-constructed prediction model to obtain the comprehensive water demand and comprehensive water-saving potential.

[0060] Among them, water demand refers to the total amount of water resources required for specific purposes (such as life, industry, agriculture, ecology, etc.) within a certain time range.

[0061] Water-saving potential is the difference between the water-saving target that can be achieved by taking various water-saving measures and the current water-saving situation, and the magnitude of this difference is the magnitude of the water-saving potential.

[0062] In this embodiment, the historical water demand and historical water-saving potential are calculated based on the historical water resource monitoring data by using the itemized quota method. The itemized quota method is a structured and systematic water resource demand analysis tool, which decomposes the water use demand into multiple specific items and quantifies each item using quota standards. The specific operation of this method is as follows:

[0063] First, collect water resource monitoring data for different time periods, including: water consumption, meteorological data, population data, and regional economic activities;

[0064] Classify the data according to the type of water use, such as agricultural water use, industrial water use, and domestic water use;

[0065] Determine the quota standards for different types of water use according to existing industry standards, local government regulations, or relevant research results; multiply the historical monitoring data by the corresponding water demand quotas to obtain the historical water demands for each type of water use; add up the historical water demands for each type of water use to get the total historical water demand;

[0066] Determine the water-saving potential for various types of water use according to industry standards or best practices, usually expressed as the percentage of water that can be saved; calculate the corresponding historical water-saving potential for each type of water use; add up the water-saving potentials for each type to get the total historical water-saving potential.

[0067] The calculation of water-saving potential is divided into three aspects: the calculation of water-saving potential for agriculture, industry, and urban domestic water use. Among them: the water-saving potential for agriculture covers three aspects: engineering water-saving, agronomic water-saving, and management water-saving; the water-saving potential for industry includes engineering water-saving, process water-saving, and management water-saving; the water-saving potential for urban domestic water use includes engineering and process water-saving.

[0068] The magnitude of water-saving potential is not only related to the current water use situation, but more importantly, it reflects the comprehensive effects of relevant factors such as water-saving models and water use structures. Water-saving potential can be expressed by the specific amount of water that can be saved or the ratio of the water that can be saved to the total water intake. In this embodiment, two representation methods are used to calculate the current water-saving potential in the study area for agricultural water use, industrial water use, and urban domestic water use.

[0069] Among them, the water-saving potential for agricultural irrigation is the amount of water consumption and water use that may be reduced by implementing certain agricultural water-saving measures at a certain stage of development through a certain investment in water-saving costs to improve the agricultural irrigation water use standard. Specifically:

[0070] W n =A0(Q n0 / μ0-Q nt / μ t )

[0071] η n =(1-Q nt μ0 / Q n0 μ t )×100

[0072] In the formula, W n (m 3 ) and η n (%) are different representation forms of the comprehensive agricultural water-saving potential considering conditions such as the rational development and utilization of water resources, water-saving in the water transmission and distribution process, water-saving in the field irrigation process, water-saving in water use management, and agronomic water-saving and yield-increasing technologies; A0 is the effective irrigation area (mu) in the current status year; Q n0 、Q nt are the average irrigation water use per mu of farmland in the current status and planned level years (m 3( / mu); mu0, mu t are the irrigation water use coefficients in the current situation and the planned water use year respectively.

[0073] The industrial water conservation potential is calculated as:

[0074] W g = Z g0 (Q g0 - Q gt )

[0075] η g = (1 - Q gt / Q g0 ) × 100

[0076] In the formula, W g (m 3 ) and η g (%) are the comprehensive industrial water conservation potential considering conditions such as industrial structure upgrading, product structure upgrading and optimization, water conservation technology transformation, and adjustment of water resource fee collection intensity; Z g0 (10,000 yuan) is the industrial added value in the current situation year; Q g0 , Q gt (m 3 / 10,000 yuan) are the water consumption per 10,000 yuan of industrial added value in the current situation and the planned year respectively.

[0077] Urban domestic water is related to the nature, scale of the city itself and the natural environment it is in, and also related to people's living styles and consumption levels. In this embodiment, engineering and technical water conservation such as increasing the popularization rate of water-saving appliances and reducing the pipe network leakage rate are considered, and the water conservation potential can be started from these two aspects.

[0078] The water conservation potential of adopting water-saving appliances is calculated using the following formula:

[0079] W czq = R0 × L0 / 1000 × 365 × J z × (P t - P0)

[0080] η czq = J z × (P t - P0)

[0081] Among them, W czs (m 3 ) and η czq (%) are the comprehensive urban domestic water conservation potential after adopting various water-saving appliances such as water-saving faucets, toilets, shower heads, washing machines, etc.; R0 (person) is the urban population in the current situation year; J z (%) is the water conservation contribution rate of water-saving appliances; L0 (L / person·d) is the urban per capita daily water consumption in the current situation year; P0 and P t(%) are the penetration rates of water-saving appliances in the current year and the planned year respectively.

[0082] The water-saving potential of reducing the leakage loss of the water supply network is calculated using the following formula:

[0083] W czg = W gw0 - W gw0 ×(1 - δ0) / (1 - δ t

[0084] η czg = 1 - (1 - δ0) / (1 - δ t )

[0085] Where, W czg (m 3 ) and η czg (%) are the water-saving potential of reducing the leakage loss of the water supply network after the transformation of the water supply network; W gw0 is the urban domestic water consumption (m 3 ) supplied by the current water treatment plant; δ0 and δ t (%) are the leakage losses of the water supply network in the current and planned years respectively.

[0086] After calculating the historical water demand and historical water-saving potential, input them into the pre-constructed prediction model. This model will learn the variation law based on this historical time series of historical water demand and historical water-saving potential, so as to predict the variation trend and obtain the comprehensive water demand and comprehensive water-saving potential.

[0087] S3. Construct a three-dimensional joint distribution model for fitting the local water production distribution, source water condition distribution, and upstream inflow distribution based on historical water resources monitoring data and water resources management information data, and generate several water supply scenarios based on the three-dimensional joint distribution model.

[0088] By analyzing historical water resources monitoring data and water resources management information data, quantitatively evaluate the overall new water resources in the target area from three aspects: local water production, source water conditions, and upstream inflows, and construct a three-dimensional joint distribution model for local water production distribution, source water condition distribution, and upstream inflow distribution.

[0089] Specifically, use historical water resources monitoring data and water resources management information data to fit the distribution of local water production; use the source water condition data and upstream inflow data in the historical water resources monitoring data to fit the source water condition distribution and upstream inflow distribution respectively; construct a three-dimensional joint distribution model according to the distributions of the three.

[0090] The three-dimensional joint distribution model contains multiple variables that affect the distribution. By adjusting the values of each variable within the corresponding range of variation, several different water resource distribution scenarios can be obtained, including local water production distribution, water source situation distribution, and upstream inflow distribution. When the water use scenario is relatively stable, the dynamic remaining amount after the increase and consumption of water resources in the target area can be estimated based on the water resource scenario and the water use scenario, thereby obtaining the water supply scenario.

[0091] S4. Input the water supply scenario, comprehensive water demand, and comprehensive water conservation potential into the pre-constructed optimal allocation model to obtain an optimized water resource allocation plan.

[0092] The optimal allocation model can be a single-objective optimal allocation model or a multi-objective optimal allocation model. There is only one evaluation function for the single-objective optimal allocation model, and only the maximum or minimum value needs to be obtained according to the specific function conditions. For example, the optimization objective of the single-objective optimal allocation model can be the shortest total water shortage period, the smallest total water shortage volume, the smallest number of water shortage times, etc. There are multiple evaluation functions for the multi-objective optimal allocation model, and the solutions using different evaluation functions are also different. That is to say, in the multi-objective optimization problem, there are multiple maximization or minimization objective functions at the same time. For example, the multi-objective optimal allocation model can be the shortest total water shortage period and the highest scheduling efficiency, etc.

[0093] According to actual needs, construct an optimal allocation model, input the water supply scenario, comprehensive water demand, and comprehensive water conservation potential into the model and solve it to obtain an optimized water resource allocation plan.

[0094] In one embodiment, the prediction model includes a sub-item trend extrapolation prediction model, a system dynamics model, and a first machine learning model. As Figure 2 shown, in step S2, inputting the historical water demand and historical water conservation potential into the pre-constructed prediction model to obtain the comprehensive water demand and comprehensive water conservation potential includes:

[0095] S21. Input the historical water demand and historical water conservation potential into the prediction model to respectively obtain the predicted water demand and predicted water conservation potential corresponding to the sub-item trend extrapolation prediction model, the system dynamics model, and the first machine learning model.

[0096] Input the historical water demand and historical water conservation potential into the sub-item trend extrapolation prediction model, the system dynamics model, and the first machine learning model in sequence to obtain the predicted water demand and predicted water conservation potential corresponding to each model.

[0097] Among them, the sub-item trend extrapolation prediction model is constructed using the trend extrapolation method. It is a prediction method that determines the predicted value by extrapolating the future based on the changing trend revealed by the historical time series of the prediction variable. The system dynamics model is a causal relationship mechanism model, which emphasizes the mutual connection and interaction between the system and the environment; its behavior pattern and characteristics are mainly determined by the internal dynamic structure and feedback mechanism of the system, and are not interfered by external factors. The quantities contained in the system dynamics model change over time, so it can be used to simulate long-term and periodic system problems. The first machine learning model completes the prediction task by learning the historical time series. Common machine learning models include LSTM (Long Short-Term Memory), XGBoost, decision trees, Transformer, etc.

[0098] S22. Based on the historical water demand and historical water-saving potential, use the first optimization algorithm to obtain the first weight values corresponding to the sub-item trend extrapolation prediction model, the system dynamics model, and the first machine learning model.

[0099] The core idea of this step is to assign different weights to the results of different models, which determines the contribution degree of each model in the final prediction result. Then, the results are weighted and summed according to these weights to obtain the final fusion result. This method is simple and easy to implement, and can flexibly adjust the weights according to actual needs to achieve the optimal fusion effect.

[0100] There is no fixed optimization algorithm for solving the first weight value of the model. The appropriate optimization algorithm can be selected according to the characteristics of the model. Common ones include the equal weight method (i.e., assigning equal weights to all models), performance-based weight allocation (i.e., assigning weights according to the performance of the model (such as accuracy, F1 score, mean square error, etc.)), confidence-based weight allocation (assigning weights according to the confidence of the model (such as variance, entropy, etc. of the prediction probability)), etc.

[0101] S23. Based on the first weight values, perform weighted fusion on the predicted water demand and predicted water-saving potential to obtain the comprehensive water demand and comprehensive water-saving potential.

[0102] Multiply the predicted water demand corresponding to the sub-item trend extrapolation prediction model by the first weight value corresponding to the sub-item trend extrapolation prediction model, multiply the predicted water demand corresponding to the system dynamics model by the first weight value corresponding to the system dynamics model, multiply the predicted water demand corresponding to the first machine learning model by the first weight value corresponding to the first machine learning model, and add the above three product results to calculate the comprehensive water demand. Similarly, the comprehensive water-saving potential is also the sum of the products of the first weight values corresponding to the three models and the corresponding predicted water-saving potential.

[0103] In this embodiment, multiple models are used to predict water demand and water-saving potential respectively. At the same time, the results of multiple models are fused with the first weight value to obtain the final result, which can avoid the limitations of a single model in multiple scenarios and increase the accuracy of the calculation results.

[0104] In one embodiment, as Figure 3 shown, S3 includes:

[0105] S31. Construct a number of water production models for simulating the water production situation in the target area based on historical water resource monitoring data and water resource management information data, and fuse the water production models to fit the local water production distribution.

[0106] The water production situation is relatively complex. Multiple water production models are fused to more accurately simulate the water production situation.

[0107] The water production models that can be adopted in this embodiment include, for example, empirical models constructed based on historical data, conceptual models constructed through a parameterization process, physical models constructed according to physical processes within the basin, models combining hydrological and ecological processes, etc.

[0108] S32. Extract source water condition data and upstream incoming water data from historical water resource monitoring data, and fit the source water condition distribution and upstream incoming water distribution according to the source water condition data and upstream incoming water data.

[0109] According to the source water condition data in the historical water resource monitoring data, fit the source water condition distribution. According to the upstream incoming water data in the historical water resource monitoring data, fit the upstream incoming water distribution. The fitting methods can include: time series analysis for processing time-correlated data, regression analysis for establishing the statistical relationship between the dependent variable (such as runoff) and one or more independent variables (such as rainfall, temperature), distribution fitting methods for describing the data distribution characteristics by selecting appropriate probability distributions, machine learning methods, etc.

[0110] S33. Synthesize the local water production distribution, source water condition distribution, and upstream incoming water distribution to construct a three-dimensional joint distribution model.

[0111] The three-dimensional joint distribution model refers to a statistical model used to describe the joint distribution relationship between three random variables, which can analyze and predict the complex relationships between multiple variables and improve the accuracy and reliability of hydrological prediction.

[0112] S34. Conduct chaotic sampling based on the three-dimensional joint distribution model to obtain a number of distribution scenarios.

[0113] Chaos Sampling generally refers to introducing a certain degree of randomness and uncertainty in the process of complex system modeling to simulate the performance of the system in different states.

[0114] First, clarify the three-dimensional joint distribution model, define the components, state variables, and possible interactions of the model. Then determine the parameters of the model and consider the potential range of variation of these parameters. The parameters may include initial conditions, control variables, etc. Next, introduce random variables or noise to simulate the inherent randomness in the system. This randomness can come from environmental noise, measurement errors, etc. Finally, use computer simulation to run the model under different random conditions and record the output results of the system, which are the distribution scenarios. In this embodiment, the more distribution scenarios, the better, and a large number of distribution scenarios can be obtained through chaotic sampling.

[0115] S35. Extract the downstream water demand from the comprehensive water demand, subtract the distribution scenarios from the downstream water demand respectively, and obtain the water supply scenarios correspondingly.

[0116] The comprehensive water demand is a prediction of the total water demand of the entire basin. Therefore, the downstream water demand can be extracted from the comprehensive water demand, and a large number of distribution scenarios are subtracted from the downstream water demand respectively to obtain a large number of water supply scenarios.

[0117] In one embodiment, the water production model includes a physically based distributed water production model, a conceptual lumped water production model, a statistical model, a rain-flood classification model, and a second machine learning model.

[0118] Among them, both the physically based distributed water production model and the conceptual lumped water production model are hydrological models. A model that uses an entity scenario to study the variation law of hydrological processes and the physical properties of certain elements is called a hydrological physical model. A conceptual basin hydrological model is a hydrological model constructed based on the physical concepts of hydrological phenomena and some empirical formulas. It generalizes the physical basis of the basin (such as the underlying surface, etc.) (such as linear reservoirs, soil layer division, storage capacity curves, etc.), and then combines hydrological empirical formulas (such as infiltration curves, unit hydrographs of runoff concentration, evapotranspiration formulas, etc.) to approximately simulate the water flow process of the basin.

[0119] The statistical model can help construct a runoff prediction model based on rainfall data. The statistical models that can be used include: linear regression models, generalized linear models, etc.

[0120] The rain-flood classification model is used to predict and classify flood events caused by rainfall.

[0121] The second machine learning model such as neural networks, random forests, support vector machines, etc. is suitable for complex non-linear relationships.

[0122] In S31, the water production models are fused, and the local water production distribution is fitted, including:

[0123] S311. Input the rainfall data of the target area into each water production model to obtain the corresponding local water production data respectively. This local water production data is the simulation data calculated by each water production model based on the rainfall data. In this embodiment, the historical local water production data can be used to evaluate the accuracy of the simulation results of the water production model.

[0124] S312. Use the bagging method to average the local water production data corresponding to each water production model to obtain a comprehensive model.

[0125] S313. Then input the rainfall data into the comprehensive model, and based on parameter uncertainty, obtain multiple runoff scenarios to fit the distribution of local water production.

[0126] The bagging method is to resample the training data to generate multiple subsets, then train multiple models and average their prediction results to fuse the models. In this embodiment, by fusing five models, the prediction results of the five models are averaged to obtain the final prediction result, so as to improve the model prediction performance and enhance the robustness and generalization ability of the model.

[0127] In one embodiment, S33 includes:

[0128] S331. Determine the Markov potential functions corresponding to the local water production distribution, source water condition distribution, and upstream inflow distribution.

[0129] S332. Transform the Markov potential function into the form of a multi-dimensional joint probability distribution function in the form of a probability distribution.

[0130] S333. Use maximum likelihood estimation to obtain several parameter groups of the Markov potential function; each parameter group estimation includes local potential function parameters, pairwise potential function parameters, high-order potential function parameters, structure parameters, regularization parameters, and constraint parameters.

[0131] S334. Based on each parameter group, construct the corresponding multi-dimensional joint probability distribution function and calculate the information criterion value corresponding to the multi-dimensional joint probability distribution function, compare the information criteria, and determine the three-dimensional joint distribution model from the multi-dimensional joint probability distribution functions according to the comparison results.

[0132] The Markov random field is a probabilistic graphical model used to model the joint probability distribution, which describes the relationship between a set of random variables. Each variable represents a node, and the edge represents the dependence relationship between nodes. The Markov random field usually uses a potential function to describe the relationship between nodes. The potential function defines the interaction between variables. By defining an appropriate potential function, the dependence relationship and constraint relationship between data can be captured.

[0133] Markov random fields can flexibly model complex data relationships, including non-linear relationships and interactions between high-dimensional data. They have local properties where the state of each node is only related to the states of its adjacent nodes. This local property can effectively reduce the computational complexity and improve the efficiency of the model when dealing with large-scale data.

[0134] Therefore, in this embodiment, a Markov random field is selected to construct a multi-dimensional joint probability distribution, which has the advantages of local properties, flexibility, structured modeling, inference efficiency, and parameter sharing compared to other methods, and can better handle data with complex structures and interdependent relationships.

[0135] In other embodiments, each variable in the multi-dimensional data is regarded as a node, and the relationship between the nodes is represented as an edge.

[0136] The Markov random field model is used to define the potential dependencies between nodes, capture the interactions and associations between data variables, and represent the joint probability distribution using a potential function. The process is as follows:

[0137] [P(X)=\frac{1}{Z}\exp{-E(X)}]

[0138] Where (X) is the value of the multi-dimensional data variable, (Z) is the normalization constant, and (E(X)) is the potential function.

[0139] Furthermore, Bayesian inference is used to optimize the parameters of the Markov random field model.

[0140] In one embodiment, as Figure 4 shown, S4 includes:

[0141] S41. Input the water supply scenario, comprehensive water demand, and comprehensive water-saving potential into the pre-constructed optimization configuration model, and use several second optimization algorithms to solve the optimization configuration model to obtain the configuration schemes corresponding to each second optimization algorithm; among them, the optimization configuration model takes the minimum total water shortage depth during the water resource scheduling period as the objective function.

[0142] S42. Obtain the historical water supply scenario based on historical water resource monitoring data, divide the historical water supply scenario, the historical water demand and historical water-saving potential extracted from the historical water supply scenario into a test data set and a validation data set, input the test data set into the optimization configuration model, use several second optimization algorithms to solve the optimization configuration model to obtain the historical configuration schemes corresponding to each second optimization algorithm, and use the validation data set to perform validation calculations on each historical configuration scheme to obtain the second weight values corresponding to each optimization algorithm.

[0143] S43. Perform weighted fusion on the configuration schemes based on the second weight values to obtain the optimized water resource configuration scheme.

[0144] The second optimization algorithm, such as the gradient descent method, linear or nonlinear programming method, genetic algorithm, Bayesian optimization algorithm, etc. Select the second optimization algorithm according to the form of the objective function, the complexity of the constraint conditions, the limitation of computing resources, and the requirements for the optimality of the solution and the computing time.

[0145] Use the selected optimization algorithm to solve the optimization configuration model respectively. Each optimization algorithm obtains the corresponding configuration scheme, and these are summarized into a set of configuration schemes.

[0146] Next, evaluate the weights of each optimization algorithm. The specific method is as follows: Divide the historical water demand and historical water-saving potential into a test data set and a validation data set. For example, take the first three-fifths of the data as the test set and the last two-fifths of the data as the validation set. Input the test data set into the optimization configuration model and use the second optimization algorithm selected in S41 to solve it, obtaining the historical configuration schemes corresponding to each second optimization algorithm. These test data sets can be used as a reference to help the model better understand the behavior of the system. Use the validation data set to verify the historical configuration schemes, calculate the performance of each second optimization algorithm, and obtain the weight values corresponding to each optimization algorithm. These weight values reflect the effectiveness of each second optimization algorithm on historical data.

[0147] In one embodiment, S1 includes: constructing a generalized river network map of the target area based on water resources management information data.

[0148] In this embodiment, the river, lake, water user, water intake, section, and node data of the target area are extracted from the water resources management information data and the rivers, lakes, water users, water intakes, sections, and nodes in the study area are generalized respectively. Based on the generalized data of the target area, use Arcgis software to construct a generalized river network map of the target area.

[0149] The steps for constructing the generalized river network map include:

[0150] Step S11: Generalize the water intakes and water users respectively based on the county administrative region, water resources division, main line, and cascade caliber. The water users and water intakes include: agriculture, industry, domestic, ecology, and ship locks.

[0151] Step S12: Process each lake into a 0-dimensional storage node based on its respective water level storage curve.

[0152] Step S13: Calculate the bottom width, bottom elevation, and slope coefficient of each river channel respectively. Use the hydraulic method to generalize the river channel based on the same passing capacity. The bottom width of the river channel is the sum of the bottom widths of the parallel generalized river channels.

[0153] Due to the dense river network and complex water flow conditions in the study area, it is difficult to collect river channel information. To reduce the workload and difficulty, in this embodiment, the river channel sections involved in the study area are generalized. On the premise of not affecting the law of river network water flow movement, the river network in the simulation calculation is merged and generalized on the basis of the natural river network and lakes. The section of the generalized river channel is a horizontal bottom slope and trapezoidal section, and the section is similar to the actual situation of the natural river network in terms of water conveyance capacity and storage capacity. Therefore, three parameters, namely the bottom width of the river, the elevation of the river bottom, and the slope coefficient, are used to describe it.

[0154] In a certain embodiment, there are two or more parallel secondary channels or tributaries. The channels that are relatively close are generalized into one channel for analysis and calculation. When these parallel channels have cross-section data and the start and end nodes are the same, the hydraulic method can also be used for generalization according to the principle of the same water passing capacity. The bottom width of the river channel is the sum of the bottom widths of the parallel generalized channels.

[0155] In another embodiment, when the parallel channels lack cross-section data and the start and end nodes are not the same, the cross-section parameters of the generalized channel are first determined by experience and then modified in the model calibration stage.

[0156] Step S14: Generalize the river channel nodes into sluice and pumping station hub nodes, water user nodes, management nodes, boundary nodes, storage nodes, and ordinary river channel nodes based on the current projects and planned projects respectively.

[0157] A hub refers to a water project facility for regulating and controlling rivers and lakes, which consists of four types of water project facilities: ship locks, sluices, pumping stations, and hydropower stations. Among them, a ship lock refers to a scheduling ship lock that consumes energy but not water, that is, after using water, the used water does not leave the river or lake where the water is taken. In this embodiment, because the structure of some hubs is relatively complex, for the convenience of calculation, the hub structure is generalized. When generalizing, the start and end two nodes are used for control. When encoding the generalized nodes of the model, between the start and end nodes, in the order of the four devices of ship lock, pumping station, sluice, and hydropower station, using the binary idea, if there is such a water conservancy project, the corresponding binary encoding character position is assigned 1, otherwise it is assigned 0.

[0158] The present invention adopts a water resources allocation method of multi-model optimization and fusion for each process to realize the quantitative assessment of water resources utilization and the improvement of the functions of the model in aspects such as hydrological simulation, scheduling, allocation, and management, which has important practical application value for alleviating the water resources contradiction in the region. The prior art over-generalizes elements such as water sources and users involved, which will make the calculation results deviate greatly from the real situation. The present invention comprehensively couples the feedback of the water-saving regulation mechanism with multiple elements such as rainfall conditions, water conditions, and project conditions, and formulates a multi-element joint real-time feedback scheduling of water project clusters.

[0159] It should be understood that although the steps in the flowcharts involved in the above-described embodiments are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear indication in this document, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-described embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be executed alternately or in turn with at least a part of other steps or steps or stages in other steps.

[0160] Based on the same inventive concept, an embodiment of the present application further provides a process-by-process multi-model optimization and fusion water resource allocation system for implementing the process-by-process multi-model optimization and fusion water resource allocation method described above. The solution provided by this system to solve the problem is similar to the solution described in the above method. Therefore, the specific limitations in one or more embodiments of the process-by-process multi-model optimization and fusion water resource allocation system provided below can refer to the limitations on the process-by-process multi-model optimization and fusion water resource allocation method in the above text, and will not be repeated here.

[0161] In one embodiment, as Figure 5 shown, a process-by-process multi-model optimization and fusion water resource allocation system is provided, including:

[0162] A collection module 501, configured to obtain historical water resource monitoring data and water resource management information data of a target area;

[0163] A prediction module 502, configured to obtain the historical water demand and historical water-saving potential of the target area based on the historical water resource monitoring data; input the historical water demand and historical water-saving potential into a pre-constructed prediction model to obtain the comprehensive water demand and comprehensive water-saving potential;

[0164] A fitting module 503, configured to construct a three-dimensional joint distribution model for fitting the local water production distribution, source water condition distribution, and upstream inflow distribution of the target area based on the historical water resource monitoring data and water resource management information data, and generate a number of water supply scenarios based on the three-dimensional joint distribution model;

[0165] An optimization module 504, configured to input the water supply scenarios, comprehensive water demand, and comprehensive water-saving potential into a pre-constructed optimal allocation model to obtain an optimized water resource allocation plan.

[0166] Each module in the above water resource allocation system with step-by-step multi-model optimization and integration can be implemented in whole or in part by software, hardware, or a combination thereof. Each of the above modules can be embedded in the processor of a computer device in hardware form or be independent of it, or can be stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to each of the above modules.

[0167] In one embodiment, a computer device is provided, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, the steps in all the above method embodiments are implemented.

[0168] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in all the above method embodiments are implemented.

[0169] In one embodiment, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the steps in all the above method embodiments are implemented.

[0170] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. And the collection, use, and processing of relevant data need to comply with the relevant laws, regulations, and standards of relevant countries and regions.

[0171] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memories. Non-volatile memories can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memories can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The databases involved in the embodiments provided in the present application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in the present application can be general-purpose processors, central processors, graphics processors, digital signal processors, programmable logic devices, data processing logics based on quantum computing, etc., without limitation.

[0172] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0173] The above-described embodiments merely represent several implementation manners of the present application. Their descriptions are relatively specific and detailed, but they should not be construed as limiting the patent scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A water resources allocation method based on multi-model optimization fusion process by process, characterized in that: The method comprises: Obtain historical water resources monitoring data and water resources management information data for the target area; Based on the historical water resources monitoring data, the historical water demand and the historical water-saving potential of the target area are obtained; the historical water demand and the historical water-saving potential are input into a pre-built prediction model to obtain the comprehensive water demand and the comprehensive water-saving potential; Based on the historical water resources monitoring data and the water resources management information data, a three-dimensional joint distribution model is constructed for fitting the local water production distribution, water source water condition distribution and upstream water supply distribution of the target area, and a plurality of water supply scenarios are generated based on the three-dimensional joint distribution model; The water supply scenario, the comprehensive water demand and the comprehensive water-saving potential are input into a pre-built optimization configuration model to obtain a water resource configuration optimization plan.

2. The method according to claim 1, characterized in that The prediction model includes a sub-item trend extension prediction model, a system dynamics model and a first machine learning model; The step of inputting the historical water demand and the historical water-saving potential into a pre-built prediction model to obtain the comprehensive water demand and the comprehensive water-saving potential includes: Input the historical water demand and the historical water-saving potential into the prediction model, and obtain the predicted water demand and predicted water-saving potential corresponding to the sub-item trend extension prediction model, the system dynamics model and the first machine learning model respectively; Based on the historical water demand and the historical water-saving potential, a first optimization algorithm is used to obtain first weight values ​​corresponding to the sub-item trend extension prediction model, the system dynamics model, and the first machine learning model; The predicted water demand and the predicted water-saving potential are weightedly fused based on the first weight value to obtain the comprehensive water demand and the comprehensive water-saving potential.

3. The method according to claim 1, characterized in that The three-dimensional joint distribution model for fitting the local water production distribution, water source water condition distribution and upstream water supply distribution of the target area is constructed based on the historical water resources monitoring data and the water resources management information data, and a number of water supply scenarios are generated based on the three-dimensional joint distribution model, including: Constructing a plurality of water production models for simulating the water production situation in the target area according to the historical water resource monitoring data and the water resource management information data, fusing the water production models, and fitting the local water production distribution; Extracting water source water condition data and upstream water inflow data from the historical water resources monitoring data, and fitting the water source water condition distribution and the upstream water inflow distribution according to the water source water condition data and the upstream water inflow data; The three-dimensional joint distribution model is constructed by comprehensively considering the local water production distribution, the water source water condition distribution and the upstream water distribution; Performing chaotic sampling based on the three-dimensional joint distribution model to obtain several distribution scenarios; The downstream water demand is extracted from the comprehensive water demand, and the distribution scenario is subtracted from the downstream water demand to obtain the corresponding water supply scenario.

4. The method according to claim 3, characterized in that The water production model includes a physics-based distributed water production model, a conceptual lumped water production model, a statistical model, a rain and flood classification model, and a second machine learning model; The fusing of the water production model to fit the local water production distribution includes: Inputting the rainfall data of the target area into each of the water production models to obtain corresponding local water production data; The local water production data corresponding to each water production model are averaged by using a bagging method to obtain a comprehensive model; The rainfall data are input into the comprehensive model, and multiple runoff scenarios are obtained based on parameter uncertainty to fit the distribution of local water production.

5. The method according to claim 3, characterized in that: The three-dimensional joint distribution model is constructed by integrating the local water production distribution, the water source water condition distribution and the upstream water distribution, including: Determine the Markov potential energy function corresponding to the local water production distribution, the water source water condition distribution and the upstream water distribution; The Markov potential energy function is converted into a multi-dimensional joint probability distribution function in the form of probability distribution; Using maximum likelihood estimation to obtain several parameter groups of the Markov potential energy function; each parameter group estimate includes local potential energy function parameters, paired potential energy function parameters, high-order potential energy function parameters, structural parameters, regularization parameters and constraint parameters; Based on each of the parameter groups, the corresponding multidimensional joint probability distribution function is constructed and the information criterion value corresponding to the multidimensional joint probability distribution function is calculated, the information criteria are compared, and the three-dimensional joint distribution model is determined from the multidimensional joint probability distribution function according to the comparison result.

6. The method according to claim 1, characterized in that The inputting the water supply scenario, the comprehensive water demand and the comprehensive water-saving potential into a pre-built optimization configuration model to obtain a water resource configuration optimization plan comprises: The water supply scenario, the comprehensive water demand and the comprehensive water-saving potential are input into a pre-constructed optimization configuration model, and the optimization configuration model is solved by using several second optimization algorithms to obtain the configuration scheme corresponding to each of the second optimization algorithms; wherein the optimization configuration model takes the minimum total water shortage depth during the water resources scheduling period as the objective function; Based on the historical water resources monitoring data, a historical water supply scenario is obtained, and the historical water demand and the historical water-saving potential extracted from the historical water supply scenario are divided into a test data set and a verification data set, and the test data set is input into the optimization configuration model, and the optimization configuration model is solved by using several second optimization algorithms to obtain the historical configuration scheme corresponding to each of the second optimization algorithms, and the verification data set is used to perform verification calculation on each of the historical configuration schemes to obtain the second weight value corresponding to each of the optimization algorithms; The configuration schemes are weightedly integrated based on the second weight value to obtain the water resource configuration optimization scheme.

7. The method according to claim 1, characterized in that The method further comprises: A river network generalization map of the target area is constructed based on the water resources management information data.

8. A water resource allocation system with multi-model optimization fusion by process, characterized in that: The system comprises: The acquisition module is used to obtain historical water resource monitoring data and water resource management information data of the target area; A prediction module, for obtaining the historical water demand and the historical water-saving potential of the target area based on the historical water resources monitoring data; inputting the historical water demand and the historical water-saving potential into a pre-built prediction model to obtain the comprehensive water demand and the comprehensive water-saving potential; A fitting module, for constructing a three-dimensional joint distribution model for fitting the local water production distribution, water source water condition distribution and upstream water supply distribution of the target area based on the historical water resources monitoring data and the water resources management information data, and generating a plurality of water supply scenarios based on the three-dimensional joint distribution model; The optimization module is used to input the water supply scenario, the comprehensive water demand and the comprehensive water-saving potential into a pre-built optimization configuration model to obtain a water resource configuration optimization plan.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images